Tidy Tuesday from 2019-07-30: Video Games Dataset

My goals

1. Use plotly instead of ggplot2 for figures.
2. No more than 3 hours from getting the data to generating the final report.

Later: repeat in Python.

Load libraries

library(dplyr)
library(plotly)

Read the data

Link to the file.

video_games <- readr::read_csv("https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2019/2019-07-30/video_games.csv")

## Parsed with column specification:
## cols(
##   number = col_double(),
##   game = col_character(),
##   release_date = col_character(),
##   price = col_double(),
##   owners = col_character(),
##   developer = col_character(),
##   publisher = col_character(),
##   average_playtime = col_double(),
##   median_playtime = col_double(),
##   metascore = col_double()
## )

Clean up the variables

release_date

Convert to date format:

video_games$release_date <- as.Date(video_games$release_date, format = "%b %d, %Y")

owners

owners is shown in N1 .. N2 format so I will assume that this is minimum and maximum number of owners. I will extract the columns and create corresponding columns with the maximum and the minimum number of players.

temp <- video_games$owners %>% stringr::str_split(., pattern = stringr::fixed(".."))
temp_min <- temp %>% sapply(., "[", 1) %>% 
  stringr::str_trim(., side = "both") %>% 
  stringr::str_replace_all(., pattern = ",", "") %>% 
  as.numeric()
temp_max <- temp %>% sapply(., "[", 2) %>% 
  stringr::str_trim(., side = "both") %>% 
  stringr::str_replace_all(., pattern = ",", "") %>%
  as.numeric()
video_games <- video_games %>% select(-owners) %>% mutate(min_owners = temp_min,
                                                          max_owners = temp_max)

game

Review missing values:

video_games %>% filter(is.na(game)) %>% 
  summarise(total = n()) %>% 
  knitr::kable(format = "html", 
               caption = "Number of rows where game name is missing",
               align = "l") %>%
  kableExtra::kable_styling()

Number of rows where game name is missing

total
3

video_games %>% filter(is.na(game)) %>% 
  knitr::kable(format = "html", caption = "Rows with missing game names",
               align = "c") %>% 
  kableExtra::kable_styling()

Rows with missing game names

number	game	release_date	price	developer	publisher	average_playtime	median_playtime	metascore	min_owners	max_owners
808	NA	2016-11-01	NA	NA	Aterdux Entertainment	0	0	NA	20000	50000
138	NA	2017-02-17	8.99	Jeroen Wimmers	Jeroen Wimmers	0	0	76	0	20000
3945	NA	2018-03-27	0.99	Paleno Games	Paleno Games	0	0	NA	0	20000

Remove missing values for games

video_games <- video_games %>% filter(!is.na(game))

developer and publisher

Are there any missing both developer and publisher columns?

video_games %>% filter(is.na(developer) & is.na(publisher)) %>% 
  summarise(total = n()) %>% 
  knitr::kable(format = "html", 
               caption = "Number of rows missing developer and publisher names",
               align = "c") %>% 
  kableExtra::kable_styling()

Number of rows missing developer and publisher names

total
47

video_games %>% filter(is.na(developer) & is.na(publisher)) %>% 
  DT::datatable()

Preview the data after the cleaning steps

Only the first 10 rows because the dataset is large, see below results of skimr::skim()

DT::datatable(video_games[1:10, ], rownames = FALSE)

Generate variable summary

skimr::skim(video_games) %>% skimr::kable()

Skim summary statistics
n obs: 26685
n variables: 11

Variable type: character

variable	missing	complete	n	min	max	empty	n_unique
developer	150	26535	26685	1	237	0	16751
game	0	26685	26685	1	212	0	26610
publisher	95	26590	26685	1	131	0	13952

Variable type: Date

variable	missing	complete	n	min	max	median	n_unique
release_date	42	26643	26685	2004-01-11	2018-12-31	2017-06-09	2942

Variable type: numeric

variable	missing	complete	n	mean	sd	p0	p25	p50	p75	p100	hist
average_playtime	9	26676	26685	9.06	117.94	0	0	0	0	5670	▇▁▁▁▁▁▁▁
max_owners	0	26685	26685	211169.95	1944091.17	20000	20000	20000	50000	2e+08	▇▁▁▁▁▁▁▁
median_playtime	12	26673	26685	5.16	68.92	0	0	0	0	3293	▇▁▁▁▁▁▁▁
metascore	23836	2849	26685	71.89	11.02	20	66	73	80	98	▁▁▁▂▅▇▆▁
min_owners	0	26685	26685	91277.87	950467	0	0	0	20000	1e+08	▇▁▁▁▁▁▁▁
number	0	26685	26685	2904.34	2377.1	1	821	2356	4524	8846	▇▅▃▃▂▂▁▁
price	3094	23591	26685	8.95	12.16	0.49	2.99	5.99	9.99	595.99	▇▁▁▁▁▁▁▁

Some game names are found multiple times:

video_games %>% group_by(game, developer, publisher) %>%
  summarise(total = n()) %>% filter(total > 1) %>%
  DT::datatable(caption = "Games listed multiple times")

video_games %>% group_by(game) %>%
  summarise(total = n(), 
            release_dates = paste(release_date, collapse = ", ")) %>% 
  filter(total > 1) %>%
  DT::datatable(caption = "Duplicated games have more than 1 release date")

Spot check for one game on Steam’s website did show different release dates. Here are the screenshots for “18 Wheels of Steel: American Long Haul” from the browsing page (release date is Jan 29, 2008)

browse

and from the game page (release date is Dec 3, 2007).

browse

Price over time

video_games %>% arrange(release_date) %>% 
  plot_ly(x = ~release_date, y = ~price, type = "scatter", mode = "markers", text = ~game) %>%
          layout(title = "Relationship between release date and the price", 
                 xaxis = list("release date", zeroline = FALSE), 
                 yaxis = list("game price, USD", zeroline = FALSE))

## Warning: Ignoring 3126 observations

Seems like there is a trend in terms of the price but not huge. We do observe a few outliers.

Developer and publishers with the most games

Select only the developers that put out at least 15 games for a smaller plot.

video_games %>% filter(!is.na(developer)) %>% 
  group_by(developer) %>% 
  summarise(total_games = n_distinct(game)) %>% 
  filter(total_games > 15) %>%
  arrange(total_games) %>%
  plot_ly(x = ~developer, y = ~total_games, type = "bar") %>%
  layout(title = "Most prolific developers",
         xaxis = list(title = list(text = ""), 
                      showspikes = TRUE, 
                      spikedash = "solid", 
                      spikecolor = "red",
                      showgrid = TRUE, 
                      categoryorder = "array", 
                      categoryarray = ~total_games,
                      tickangle = 45), 
         yaxis = list(title = list(text = "games"))) 

Similar approach to the publishers, but change the game minimum to 50.

video_games %>% filter(!is.na(publisher)) %>% 
  group_by(publisher) %>% 
  summarise(total_games = n_distinct(game)) %>% 
  filter(total_games > 50) %>%
  arrange(total_games) %>%
  plot_ly(x = ~publisher, y = ~total_games, type = "bar") %>%
  layout(title = "Most prolific publishers",
         xaxis = list(title = list(text = ""), 
                      showspikes = TRUE, 
                      spikedash = "solid", 
                      spikecolor = "red",
                      showgrid = TRUE, 
                      categoryorder = "array", 
                      categoryarray = ~total_games,
                      tickangle = 45), 
         yaxis = list(title = list(text = "games"))) 

Hey! Big Fish Games!

How many developers are also publishers and how many games do they have?

video_games %>% filter(!is.na(developer) & !is.na(publisher)) %>% 
  filter(developer == publisher) %>% 
  group_by(publisher) %>% 
  summarise(total_games = n_distinct(game)) %>% 
  filter(total_games > 15) %>%
  arrange(total_games) %>%
  plot_ly(x = ~publisher, y = ~total_games, type = "bar") %>%
  layout(title = "Most prolific devlopers who are also publishers",
         xaxis = list(title = list(text = ""), 
                      showspikes = TRUE, 
                      spikedash = "solid", 
                      spikecolor = "red",
                      showgrid = TRUE, 
                      categoryorder = "array", 
                      categoryarray = ~total_games,
                      tickangle = 45), 
         yaxis = list(title = list(text = "games")))

Most owned games

Ownership will be considered as the max number of owners.

video_games %>% arrange(desc(max_owners)) %>% slice(1:100) %>% 
  plot_ly(x = ~game) %>% 
  add_trace(y = ~max_owners, name = "max owners", type = "scatter", mode = "markers") %>%
  add_trace(y = ~min_owners, name = "min owners", type = "scatter", mode = "markers") %>%
  layout(title = "Top 100 most owned games",
         xaxis = list(title = list(text = ""),
                      tickangle = 45),
         yaxis = list(side = "left", title = "owners"))

Dota2 is the clear winner. My husband commented that it is also a free to play game which will definitely contribute the the overall number of people owning it.

Are the most owned games also have the highest scores?

video_games %>% arrange(desc(max_owners)) %>% slice(1:100) %>% 
  plot_ly(x = ~max_owners, y = ~metascore, text = ~game, type = "scatter", mode = "markers") %>% 
  layout(title = "Top 100 most owned games and their metascore",
         xaxis = list(title = "maximum estimated number of owners"),
         yaxis = list(title = "metascore"))

## Warning: Ignoring 30 observations

Not necessarily! Again, tapping into my husband’s knowledge of video games, he explained that some of the games weren’t available through Steam right away.

Many hardcore players?

Compare average and median time played for each game in the last two weeks (do remove the missing values and 0)

video_games %>% filter(!is.na(average_playtime) & !is.na(median_playtime) & 
                         average_playtime != 0 & median_playtime != 0) %>%
  plot_ly(x = ~average_playtime, y = ~median_playtime, 
          type = "scatter", 
          mode = "markers", 
          hoverinfo = "text", 
          text = ~paste0("average: ", average_playtime, 
                                             "\nmedian: ", median_playtime,
                                             "\ngame: ", game)) %>% 
  layout(title = "Average vs median playtime",
         xaxis = list(title = "average over two weeks, min"),
         yaxis = list(title = "median over two weeks, min"),
         shapes = list(type = "line", x0 = 0, y0 = 0, x1 = 1, y1 = 1, 
                       yref = "paper", xref = "paper", 
                       line = list(color = "red")))

The points below the diagonal line have average larger than median which means that a lot more people don’t spend too much time playing the game; the points above the diagonal line have median larger than the average which means that those games have a few very hard core players (they spend a lot of time playing the game).

Conclusions

I have certainly not explored all information in the data. Of course, it would be worthwhile looking at the date of release and the number of minutes playes. Are older games still holding up?

From conversation with my husband it seems that it is valuable to have some domain knowledge about this area and therefore additional variables. For example, information about where the game was realeased first (console?), whether the game is free to play and how many players play a certain game.

For this first time I definitely spent less time had I used ggplot2 library instead of plotly, however, this wasn’t the goal.

Problems with plotly

I have observed a strange behavior with plotly when I was trying to plot the top 100 games owned by most people and order the x axis by the maximum number of owners:

video_games %>% arrange(desc(max_owners)) %>% slice(1:100) %>% 
  plot_ly(x = ~game) %>% 
  add_trace(y = ~max_owners, 
            name = "max owners", 
            type = "scatter", 
            mode = "markers") %>%
  add_trace(y = ~min_owners, 
            name = "min owners", 
            type = "scatter",
            mode = "markers") %>%
  layout(title = "Top 100 most owned games",
         xaxis = list(title = list(text = ""),
                      tickangle = 45, 
                      categoryorder = "array", 
                      categoryarray = ~ max_owners),
         yaxis = list(side = "left", title = "owners"))

Why in this case the x axis labels are cut off and the first 2 values in the x axis are 200,000,000? I also seem to replicate the bahavior when I use categoryarray = ~min_owners, but then I have 100,000,000. I think this is a reproducible example that can be posted as a question on stack overflow.